Prove Your Controls,
Don't Just Log Them
We don't have an AI governance problem. We have an AI evidence problem.
Organizations routinely maintain policies, risk registers, model cards, and operational telemetry. Yet when it comes to high-stakes AI deployment, auditability remains fragmented, weak, and difficult to verify as a coherent chain of accountability.
The Observability Gap
Ordinary logs are mutable, unstructured, and optimized for operations—not custody. They cannot prove what rules were applied.
The Policy Gap
GRC platforms document policies, but they cannot prove that those policies actively constrained runtime behavior at the moment of execution.
What is cognitiveinsight.ai?
Welcome to the reference portal for the AI Governance and Evidence Infrastructure (AGEI). AGEI provides an open reference infrastructure for collecting, linking, and verifying AI-governance evidence across seven domains: models, agents, human in the loop, policy and delegated authority, evidence and assurance, evidence vault, provenance and transparency, Shadow AI, and privacy-aware data governance.
Agentic AI Governance Managing Delegated Authority through five planes of governance: identity, policy, privilege, execution, and evidence.
Our goal is not to impose a rigid rulebook, but to share years of research, design patterns, and a complete relational schema database contract so that the developer and security communities can collectively build a common, lightweight, and non-cumbersome way to implement controls and verify audit trails.
All concepts, documentation, and database contracts on this site are 100% open-source, licensed under the Apache License 2.0.
The Core Blueprint
1. The "Proof, Not Logs" Philosophy
Capture lightweight, cryptographically signed "footprints" continuously, and only materialize detailed evidence "capsules" when triggered.
Read the Philosophy Guide →2. The Reference Architecture
An 8-layer framework that maps high-level Policy Intent down to executable Gate Enforcement and a Tamper-Evident Evidence Vault.
Explore the Architecture →3. Guarding Runtime Actions
Govern authority, not just capability. Five Governance Planes and Pre-Action Proof-Carrying Tool Execution.
Learn About Agent Governance →4. The Database Schema Contract
A fully realized, 60-table relational schema blueprint organizing organizational boundaries, rules, receipts, and extensions.
Explore the Schema Reference →